The manager ratings model that I introduced last month has been improved over the past few weeks according to feedback I received as well as discussions I’ve had with a few football nerds. The goal was to enhance the relativity of the model, to better reflect both long-term and short-term expectations. Some of the changes are:
The most common question I had received was the absence of negative rating. The basic idea behind not giving a negative rating initially was the fact that clubs don’t get negative points for a loss. But not all losses are made the same, just like not all wins are equal. Discounting the magnitude of a specific loss tends to become unfair — case in point: Pep Guardiola would get the same rating for a loss against Leicester as a loss against Chelsea. In the improved model, the rating received for Guardiola for those losses are -2.76 and -1.46 respectively.
Net transfer spend
The second most common feedback has been about the high weighting on net transfer spend. Although I have not changed the weighting on net transfer spend, I have changed the way it is calculated. The main issue with net transfer spend is that it does not have high correlation with fan’s expectations. Expectations are largely set by transfer fees paid, and are only partially attenuated by transfer fees received. For example, Liverpool had zero net transfer spend by selling fringe players like Christian Benteke and Joe Allen and buying first team starters like Sadio Mane and Georginio Wijnaldum — but fans’ expectations have not remained even after those transactions. Along this line of thinking, I have reduced the weighting for outgoing players and increased the weighting for incoming players. This ensures that a team that buys and sells four players for zero net spend are expected to do better than a team that didn’t make any transfers.
One key factor, which I should have never ignored was considering the historical performance of teams. In my previous model, the expectations were only based on last season’s performances. So even if Chelsea were to finish ninth this season, it would look like an improvement from last season, but chances are we would still consider it a massive disappointment because of Chelsea’s Premier League history. Chelsea’s current first place position is a massive improvement on last season, but in terms historical expectations, it’s a bit less surprising. The improved model considers the historical performance of teams as well, however, a higher weighting is given to performance of last season.
Increased emphasis on form
This one I felt was necessary as managers were not getting enough credit for beating the in-form teams. One prominent example I can think of was when Spurs beat Manchester City. City were on the back on six consecutive wins and were the overwhelming favorites going into the fixture. However, Spurs played the game of the season and beat the title favorites 2-0. This victory also stared the downfall of Manchester City as Guardiola only managed 21 points in the next 13 fixtures (Stoke and West Ham had 19, while Chelsea had 39). The improved model gives Mauricio Pochettino 11.5% higher rating points for victory in that match.
Considering these changes, the updated ratings are:
When we last looked at these ratings, Jürgen Klopp was the runaway leader. While he still remains top, the difference between Klopp and Conte has significantly reduced in this model from 4.208 to 0.147.
One thing that should be noted here is that Chelsea are now getting slightly lower returns on wins compared to the previous model with the historic considerations added. In recent history, Chelsea have done better than Liverpool due to which the expectations of Chelsea fans are bound to be higher than Liverpool fans
Another major factor keeping Klopp in the lead has been Liverpool’s performance against the top six. The team’s performances in these matches give higher positive ratings in case of a win and higher negative ratings in case of a loss. Klopp has won three and drawn two, while Conte has won three and lost two. One of the two losses were against Liverpool and at Stamford Bridge, which gave Klopp a rating of 4.85 and Conte -2.04. This fixture alone has given Klopp 6.89 more rating points than Conte.
The biggest winner of the updated model has been Mauricio Pochettino whose relative rating has increased by 11.50%. Pochettino is the only manager without a negative rating this season, meaning that Spurs haven’t been upset once yet this season. Their two losses came at Chelsea and at Manchester United, games they were probably expected to lose. Pochettino also gets the highest returns on wins compared to other top six teams, which help him greatly in his overall rating.
As you can see the top three of Jürgen Klopp (2.097), Antonio Conte (2.005) and Mauricio Pochettino (1.976) are the runaway leaders. These three have distanced themselves from the rest of the pack with future Arsenal manager Eddie Howe (1.662) in fourth place. The gap is significant enough for these three to already be considered for manager of the year nominations, come May.
Next up: Why Conte will do what no manager in history of Premier League has done.